On Fri, Jun 24, 2005 at 10:55:33PM +0100, Peter C. Tribble wrote: > I was just looking at the observability community page, at > http://www.opensolaris.org/os/community/observability/ and it set > off a few thoughts about that page, some of the tools, and some things > that may be missing.
Cool, good to hear from you! > I actually much prefer top to prstat. Comparing the two, prstat is missing > a few things that I use all the time in top - the CPU states, the > memory/swap statistics, and the last pid. Furthermore, you can change the > sort order and filter by user on the fly. And generally, I just find top > easier to read. Well, to each his own ;-) Sounds like a batch of RFEs for prstat. Too bad Andrei's out of town for a while. > The page describes prtconf as x86 only. Strange! It works on all my > sparc machines... Yep, that's blatantly wrong. > Alongside prtconf and prtdiag, I think mention should be made of prtfru and > prtpicl. Definitely. > I wrote a java kstat browser, and I'm tempted to do the same for prtpicl > (and maybe some of the other prt* commands), to make it easier to see the > hierarchy and drill down into it. It's hard to wade through prtconf and > prtpicl output (especially with -v on). Sounds cool. If you send me your opensolaris username I can add you as a community leader and you can create a page for the tool. Or we can put a link if you have it hosted somewhere already. I was thinking along the lines of an "unbundled" section we we can place cool extra tools. > I see no mention of busstat. (Of particular use might be a mention of the > sort of questions it might provide answers to.) I've never used it myself, but hopefully someone with more experience can add some info. > One area conspicuous by its absence on the page is network observability. > And some of the tools we used to have no longer work (or work properly). > For example, nfswatch used to be a very useful tool, and no longer works > as it hasn't been updated. Another tool we used to use extensively to view > network traffic was etherman. There's obviously the use of netstat and > nfsstat. Yeah, that's a very good point. I'll add links to the basic tools (snoop, netstat, nfsstat). One of the items on our very long list of things to do is an NFS DTrace provider and a Network DTrace provider, but they're rather complicated (particularly the network one). But with those in place could actually build much more useful observability tools. > One thing I would like to see much more of is graphical display of what's > going on. You can have all the numbers in the world, but good graphical > visualization tools can make it so obvious what's going on! Yes, this is very true. Our main effort in this area is a Java JNI binding for libdtrace. With this in place, people will be able to rapidbly deploy GUI applications built on top of DTrace. This interface is nearly finished, and should be part of OpenSolaris real soon now. While is hard to write "The DTrace GUI", having a proper Java interface will let us test out alternative visualization tools. Thanks for all the feedback. I'll make some of the obvious changes, but if you want to help lead the community I'd be more than happy to add you to the list. - Eric -- Eric Schrock, Solaris Kernel Development.
